A fast approach for fusion of hyperspectral images through redundancy elimination

  • Authors:
  • Ketan Kotwal;Subhasis Chaudhuri

  • Affiliations:
  • Indian Institute of Technology Bombay, Mumbai, India;Indian Institute of Technology Bombay, Mumbai, India

  • Venue:
  • Proceedings of the Seventh Indian Conference on Computer Vision, Graphics and Image Processing
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

The fusion of hyperspectral images is an important area in research and applications. Several fusion techniques have been developed in the literature for visualization of hyper-spectral data. The amount of computation needed for such techniques is directly related to the volume of the data. Most of these techniques involve a significant amount of computation due to high volume of the data, making the fusion processes slow. We analyze the statistical characteristics of this data in order to develop a technique for faster fusion. The image bands in the hyperspectral data represent the response of the scene collected over contiguous narrow bands of wavelength. The adjacent bands being captured over neighboring wavelength bands, these images exhibit a very high degree of similarity. The fusion of these adjacent image bands, thus adds a very little amount of additional information. We exploit this redundancy in the data to provide a novel scheme for rapid visualization. We propose a scheme for the selection of a subset of images from a hyper-spectral image cube that can produce fusion results with a very small amount of degradation in the quality compared to the quality of the result using the same technique of fusion applied over the entire data.